Thesis
Novel workflow for driving a 3D face rig from a reference video using deepfakes
Master of Science (M.S.), Drexel University
Jun 2021
DOI:
https://doi.org/10.17918/00000748
Abstract
There have been many advances in technology over the years when it comes to animation. We have seen the industry transition from 2D animations from the 40s like Mickey Mouse to 3D animations like Toy Story. However, one of the issues that the industry still hasn't been able to resolve is recreating and animating a human face in a 3D environment without falling into the "uncanny valley". With this research, I propose a novel way to animate a 3D face using the deepfake technology. I started by initially training the deepfake model using real life footage of my face. I then created a realistic 3D face and used motion tracking technique to animate the 3D face using motion capture (mocap) in Blender. Moving forward, the pre-trained deepfake algorithm was used to drive the animation of the 3D head, and finally I compared the mocap and the deepfake animations by conducting a survey. There were a few areas where one of the animations was perceived favourably over the other, however the overall perception of the animations was similar on the majority of the factors, which can be attributed to a few limitations mentioned in the paper.
Metrics
59 File views/ downloads
35 Record Views
Details
- Title
- Novel workflow for driving a 3D face rig from a reference video using deepfakes
- Creators
- Prnav Sood
- Contributors
- David A. Mauriello (Advisor)
- Awarding Institution
- Drexel University
- Degree Awarded
- Master of Science (M.S.)
- Publisher
- Drexel University; Philadelphia, Pennsylvania
- Number of pages
- viii, 29 pages
- Resource Type
- Thesis
- Language
- English
- Academic Unit
- Digital Media; Drexel University; Antoinette Westphal College of Media Arts and Design
- Other Identifier
- 991015080649204721